Validation and automation of the attention bias test for anxious states in sheep (AEC16/19)

Assessment of emotional states is becoming an increasingly important part of animal welfare research, but emotional state is hard to measure and often requires time consuming or complicated tests. A threat perception test has been developed as a measure of anxiety in sheep and validated using pharmacological models of anxiety. While it appears that the responses we measure in the test are directed towards the threat of a dog, a controlled study had not been conducted to confirm this. The main objective of this study was therefore to further investigate the behavioural responses of sheep in the threat perception test to differentiate between responses to the dog versus responses to the novel testing environment itself. A secondary aim of this study was to automate some of the behavioural measures taken during the test. The collection of key threat perception measures (vigilance and attention to threat) from videos is a long and labour intensive process. Accelerometers or similar devices have been used previously on sheep, attached via halters or collars, to monitor animal movements and feeding behaviours. This study investigated the use such devices to automate the collection of vigilance and attention to threat data, making the test faster, more practical and more accurate. Importantly, this study aimed to determine whether the attachment of data loggers to sheep would alter the behaviour of the animals during testing.